Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
China’s attempts to dominate the world of artificial intelligence could be paying off, as industry experts and tech analysts tell CNBC that Chinese AI models are already very popular and are maintaining the pace, and even surpass, those of the US in terms of performance.
AI has become the latest battleground between the United States and China, and both sides consider it a strategic technology. Washington continues to restrict China’s access to cutting-edge chips designed to help power artificial intelligence amid fears the technology could threaten US national security.
It has led China to pursue its own approach to increasing the appeal and performance of its AI models, including relying on open-source technology and developing its own super-fast software and chips.
China is creating popular LLMs
Like some of the top US companies in the space, Chinese AI firms are developing so-called large language models, or LLMs, which are trained on large amounts of data and underpin applications such as chatbots.
Unlike the OpenAI models that power the wildly popular ChatGPT, however, many of these Chinese companies are open source or open weight LLM development which developers can download and build on top of for free and without strict licensing requirements from the inventor.
On Hugging Face, an LLM repository, Chinese LLMs are the most downloaded, according to Tiezhen Wang, a machine learning engineer at the company. Qwen, a family of AI models created by the Chinese e-commerce giant Alibabais most popular in Hugging Face, he said.
“Qwen is rapidly gaining popularity due to its excellent performance on competitive benchmarks,” Wang told CNBC via email.
He added that Qwen has a “very favorable licensing model”, which means it can be used by companies without the need for “extensive legal reviews”.
Qwen comes in different sizes or parameters as it is known in the LLM world. Large parameter models are more powerful but have higher computational costs, while smaller ones are cheaper to run.
“Regardless of which size you choose, the Qwen is likely to be one of the best performing models available right now,” added Wang.
DeepSeek, a start-up, also recently came up with a model called the DeepSeek-R1. DeepSeek said last month that its R1 model competes with OpenAI’s o1, a model designed to reason or solve more complex tasks.
These companies claim that their models can compete with other open source offerings such as Goal‘s Llama, as well as closed LLMs like OpenAI’s, through various functions.
“Over the past year, we’ve seen the rise of Chinese open source contributions to AI with very strong performance, low cost of service and high performance,” Lux Capital partner Grace Isford told CNBC by e-mail.
China pushes open source to go global
Open sourcing a technology serves several purposes, including driving innovation as more developers gain access to it, as well as building a community around a product.
It’s not just Chinese companies that have launched open source LLMs. Facebook parent Meta, as well as European start-up Mistral, also have open-source versions of AI models.
But with the tech industry caught in the crosshairs of the geopolitical battle between Washington and Beijing, open source LLMs offer Chinese companies another advantage: allowing their models to be used globally.
“Chinese companies want their models to be used outside of China, so this is definitely a way for companies to become global players in the AI space,” said Paul Triolo, partner at global advisory firm DGA Group, to CNBC via email.
While the focus right now is on AI models, there is also a debate about what applications will be built on top of them and who will dominate this global Internet landscape in the future.
“If you assume that these frontier-based AI models are table stakes, it’s about what these models are used for, like accelerating frontier engineering science and technology,” Lux Capital’s Isford said .
Current AI models have been compared to operating systems, such as from Microsoft windows, Googleis android and AppleiOS, with the potential to dominate a market, as these companies do in mobile and PC.
If true, this raises the stakes for building a dominant LLM.
China is focusing on large language models (LLM) in the artificial intelligence space.
Blackdovfx | Stock | Getty Images
“They (Chinese companies) perceive LLMs as the center of future technology ecosystems,” Xin Sun, senior professor of Chinese and East Asian business at King’s College London, told CNBC via email.
“Their future business models will depend on developers joining their ecosystems, developing new applications based on LLMs, and attracting users and data from which profits can subsequently be generated through various means, including but not limited to further, drive users to use their cloud services.” Sun added.
Chip restrictions cast doubt on China’s AI future
AI models are trained on large amounts of data, which require large amounts of computing power. Currently, Nvidia is the lead designer of the chips needed for this, known as graphics processing units (GPUs).
Most major AI companies are training their systems on Nvidia’s highest-performance chips, but not in China.
Over the past year or so, the US has increased export restrictions on advanced chip and semiconductor manufacturing equipment to China. It means NvidiaCutting-edge chips cannot be exported to the country, and the company has had to create semiconductors that meet export sanctions.
Despite these limitations, however, Chinese companies have still managed to launch advanced AI models.
“The major Chinese technology platforms currently have enough access to computing power to continue to improve models. This is because they have stockpiled a large number of Nvidia GPUs and are also taking advantage of domestic GPUs from Huawei and other companies,” said Triolo of DGA group.
Indeed, Chinese companies have been drive efforts to create viable alternatives to Nvidia. Huawei has been one of the main players in the pursuit of this goal in China, while companies like it Baidu and Alibaba have also been investing in semiconductor design.
“However, the gap in terms of advanced hardware computing will widen over time, especially next year, as Nvidia rolls out its Blackwell-based systems that are restricted for export to China,” Triolo said.
Lux Capital’s Isford noted that China has been “systematically investing in and growing its entire domestic AI infrastructure stack outside of Nvidia with high-performance AI chips from companies like Baidu.”
“Whether or not Nvidia chips are banned in China will not stop China from investing and building its own infrastructure to build and train AI models,” he added.
